Proposal for an industrial monitoring architecture oriented to Industry 4.0

Abstract

In the work, an industrial monitoring architecture is proposed, based on concepts of the so-called fourth industrial revolution or Industry 4.0. It is organized in a modular way, to facilitate its use and guarantee its scalability. The design of the architecture had, as a premise, to guarantee the requirements of lightness, open code and use of artificial intelligence tools. For achieving the first requirement, MQTT was chosen, as it is a lightweight message protocol. For their part, all the tools and code libraries used are permissive and (except one of them) compatible with the Debian free software guidelines. Finally, the conception of the modeling module guarantees the possibility of using various artificial intelligence tools to perform classifications and regressions, which allow the indirect monitoring of variables. For a preliminary validation of the system, it has been deployed in a monitoring system of the equatorial welded joint dimensions in 10 kg liquefied gas cylinders. These dimensions are determined, indirectly, through the digital processing of the captured images, by using a convolutional neural network. The deployed system widely showed its capability to fulfill the monitoring task for which it was designed.